A small literature explores the association of risk preferences with health behaviors

Reproductive health researchers and practitioners have documented that perceptions of the risk of pregnancy and of STIs shape sexual and reproductive behavior in addition to pregnancy intention and access to contraceptive methods. These perceptions and behaviors are partly the result of individual tolerance for risk and uncertainty. Tversky and Kahneman instigated a robust area of inquiry by first describing how risk preferences shape choices and behavior. These same theories of how humans make decisions under uncertainty may shed light on differences in reproductive outcomes. Economists concerned with explaining behavior, including wealth accumulation, focus on measures of economic preferences that govern differences in decision-making. These preferences, including the willingness to assume risk, strongly predict financial behavior and outcomes. Analyses that examine how risk preferences may extend beyond financial behavior to explain behavior in other domains, including health, appear less frequently in the literature. A particularly unexplored area of research concerns how risk preferences affect decision-making surrounding reproduction, contraceptive behavior, and sexual risk taking, where intention-behavior inconsistencies are widely acknowledged. This paper tests the hypothesis that the propensity to take risk manifests not only in the financial decisions commonly studied by decision theorists, but also in less-well studied choices that affect reproductive health. I explore whether individual measures of financial risk tolerance predict important reproductive outcomes of sexual and contraceptive behavior using data from the 1997 cohort of the National Longitudinal Survey of Youth . National and international public health priorities include reduction of unintended pregnancy. Defined as pregnancies that are mistimed or unwanted, unintended pregnancies comprised 45% of all pregnancies to women aged 15-44 in the United States in 2011. Half of those pregnancies ended in abortion.

Epidemiologic literature, moreover,cannabis growing supplies reports associations between unintended pregnancy and negative health and mental health outcomes for mothers and children. Women who experience unintended pregnancies also are more likely to report perceived stress, low social support, and depressive feelings.Evidence further suggests a connection between unintended pregnancy and risk behaviors during pregnancy, such as smoking and alcohol drinking. It is estimated that unintended pregnancies cost the US more than $20 billion per year in expenses for births, abortions and miscarriages. Unintended pregnancies, like many health outcomes, are differentially experienced in the population. They concentrate among women of color and low-income women. Young age, low education, previous pregnancies, non-married status, and living in an urban neighborhood, explain some but not all of the concentration of unintended pregnancy among poor and minority women. Nonuse or inconsistent use of contraception is common among women at risk of unintended pregnancy, meaning those sexually active with a stated desire not to get pregnant.Reproductive health literature reports inconsistencies between intention and behavior regarding contraceptive use and pregnancy. A California study following young women who initiated a new method of contraception and who reported not wanting to be pregnant within a year found high rates of discontinuation over the year. An analysis of a nationally representative dataset found that 25% of non Hispanic black women, 16% of Hispanic women, and 14% of non-Hispanic white women did not use contraception despite risk of unintended pregnancy. Additional studies have shown disparities in contraception use for non-Hispanic black and Hispanic women. These differences may result, at least in part, from lack of information about method availability, especially long acting methods. Research on contraceptive attitudes has reported fear of side effects and mismatches of desired method features to selected method features to be a reason for nonuse. Other research suggests that provider biases may also lead to disparities in information: in one such study, providers of contraceptive counseling recommended IUDs to low-income women of color more often than to white women.

Misperception of pregnancy risk may also result in reduced contraceptive use and subsequent unintended pregnancy regardless of knowledge of available methods. A study of family planning patients found, for example, that underestimation of the likelihood of conception predicted unprotected sex. Foster et al. found that nearly 46% of women engaged in unprotected intercourse in the past three months underestimated the risk of conception.A study of women seeking abortion services found that the majority of women had an inaccurately estimated the risk of pregnancy prior to conception. Structural and psychosocial factors also affect unintended pregnancy risk. Relationship factors, including reproductive coercion, drive contraceptive decision-making and ability to use contraception even in contexts where pregnancy is not desired. A qualitative study exploring determinants of inconsistent use found that eroticism of unprotected sex and the risk of conception was a powerful explanatory factor. Additionally many women, particularly young, poor, and uninsured women, lack access to reproductive healthcare. Micro economists have long attempted to explain differences in choices given equal information. Much of this work falls under the rubric of “behavioral economics,” the study of decisional biases and preferences. Primary questions in the field include: what leads people to behave in ways inconsistent with intentions? Given that differences in information alone unlikely explain differences among groups in unintended pregnancy, looking to behavioral economics for suggestions of other determinants seems warranted. Risk preferences, frequently studied in social science, are strong determinants of financial decision-making, including investment and savings. While elicitation methods vary, a common approach to assess preferences includes a series of questions assessing willingness to take gambles with lifetime income. People willing to take fewer gambles are generally deemed risk averse and those more willing to take gambles are risk tolerant. Behavioral economists have argued that risk preferences in humans arise from loss aversion or the tendency to risk more to avoid a loss than to realize a gain even if the prospective losses and gains are equal.

The argument has garnered credibility in comparison to a standard economic model, which assumes “rational agents” who make choices to maximize utility and who would, therefore, exhibit indifference in choices between equal loss and gain.One seminal study found that a measure of financial risk tolerance predicted risk behavior including smoking and drinking. Barky validated the now widely used measure of risk preferences in a nationally representative sample in the Health and Retirement Study . Risk aversion has also been linked to cancer screening behavior , smoking, heavy drinking, obesity, and non-use of a seat belt. Despite results from the above studies, and the widely-held belief that loss aversion and risk preferences affect choices under uncertainty, surprisingly little attention has been paid to these measures as determinants of sexual and reproductive behavior. In the best example of this limited literature, Schmidt hypothesized that when the risk of pregnancy appears highly uncertain, risk preferences help explain variability in timing of childbearing . She found risk tolerance correlated with earlier childbearing at young ages and earlier timing of marriage. She suggested that the association between early childbearing and increased risk tolerance may be a product of less contraceptive use, although she did not directly test this connection. With exception, much work argues that risk preferences vary with age, income and gender. Gender differences in risk aversion, with females expressing greater aversion, have been hypothesized to arise from different reproductive investment strategies. A few studies have shown that risk tolerant women more likely delay marriage,cannabis indoor growing indicating the apparent importance of risk tolerance to demographic behavior. Schmidt found that the effect of risk tolerance on fertility timing varied by marital status, such that for both married and unmarried women, higher risk tolerance predicted early birth at young ages. Among married women it was also associated with delayed fertility later in life. In this paper, I ask whether propensity to risk-taking in financial decisions affects reproductive behavior. Using data from the 1997 cohort of the National Longitudinal Survey of Youth , I examine several outcomes relevant to reproductive health: sex with high-risk partners, number of sex partners, consistency of contraception use, and effectiveness of contraceptive method. I conduct stratified analyses for contraceptive consistency and effective method use among unmarried and non-cohabitating women and among married or cohabitating women. Number of sexual partners. Respondents were asked the number of partners they had sexual intercourse with since the last interview. As this variable was highly right-skewed, I truncated responses above five to greater than or equal to five partners. Consistent contraceptive use. I derived a measure of consistency of contraceptive use from several questions in the survey. Respondents are first asked the number of times they had sex and then the number of times they used a condom or other birth control. If they could not recall the number, they were asked the proportion of the time that they used a method.

Combining the frequency of intercourse and condom or birth control use questions allowed me to create a percent condom or birth control variable. This is variable categorized into nonuse , inconsistent use , and perfect consistent use . I recoded anyone that reported sterilization as a consistent user even if they reported a lower birth control usage percent. Contraceptive method type. Respondents were asked: “ thinking of all the times that you have had sexual intercourse since the last interview, how many of those times did you or your sexual partner or partners use a condom or female condom?” if they replied 1 or greater, I coded them as using condoms. The following question assessed additional contraceptive use asking which “one of these methods did you or your partner use most often, either with or without a condom or female condom?” Respondents could select one of the following: withdrawal , rhythm , spermicide , diaphragm , IUD , morning after pill , birth control pills, Depo-Provera or injectables, Norplant, patch or ring , cap or shield , had vasectomy or tubal ligation, or no other method. Since the questions were asked separately, condom use could be in addition to or in absence of other contraception use. Contraceptive method effectiveness. I grouped contraceptive methods together based on typical use pregnancy prevention effectiveness rates using categories of low effectiveness , medium effectiveness , and high effectiveness . Participants were coded as medium effectiveness with condoms only if condoms were reported as the only method. Confounding variables were selected a priori based on literature as characteristics that would affect both risk aversion and sexual behavior and included: age, race/ethnicity, education, religion, marital status, parity, insurance, poverty. Age serves as an important control variable, as both risk aversion and fertility intention and contraceptive behavior change with age. Age was assigned at year of risk preference measurement and ranges from 26-32. Race/ethnicity information was recorded at entry into the cohort in 1997. Respondents self-reported race ethnicity, which I divided into Hispanic, non-Hispanic white, non-Hispanic black, and mixed race/other. For multivariable models, I dropped the mixed race/other respondents due to positivity concerns. Education influences risk perception as well as fertility timing and method choice. Risk aversion has been shown to decrease with later schooling and age. I categorize educational attainment into the following categories: some high school, high school degree, some college, and college degree or higher. In addition to age, reproductive history is an important predictor of current reproductive behavior. I therefore include a control for parity . Unintended pregnancy may carry different meaning in the context of being partnered and not partnered. While there is evidence that partnership may affect risk preference and also sexual behavior, I would not expect a different relationship of risk preferences to high risk sex by partnership status. Partnership status can be thought of as an effect modifier to the relationship between risk preferences and contraceptive behavior. Additionally relationship context and power may influence ability to use contraception. Marital status is categorized as: never married/not cohabitating, never married/cohabitating, currently married, and widowed, separated or divorced. While each wave of NLSY97 contains detailed information on contraceptive method use and type, no wave measures pregnancy intention. I cannot, therefore, attribute nonuse to wanting pregnancy or use to wanting to avoid pregnancy. I attempt to indirectly assess intention through fertility expectations. First I include a measure of fertility expectations assessed with the following question: “In five years, what is the percent chance that you will have child?” Respondents report a range of 0-100%. I code ‘don’t know’ responses as 50%. While expectations may differ from intention, they have been shown to access the same construct .